#🧵 how should the CLI print objects?

1 messages · Page 1 of 1 (latest)

obsidian owl
#

roger, roger

#

if we can get away with it, I like the idea of printing nothing, but there's a wrinkle imo in shell: if we print nothing, that makes e.g. foo $(container) a little difficult to reason about, because there's no tangible value you can imagine being passed to foo

#

another strawman: just print something like Container@xxh3:abcdef? it's short, and lets you identify whether an object is the same after repeated calls

inland birch
#

oh like the digest?

#

digest form of the ID?

#

Yeah I love that. Actually would be good to train users to form a mental model of content-addressed immutable objects

#

That's the sentenc eI keep repeating in demos when I need to explain Dagger in 30 seconds... "A DAG of immutable objects transformed by repeatable functions"

#

I also start by describing it as a programmable build engine. Then -> DAG of immutable objects. Then -> by the way LLMs work like builds. You don't "run" an agent you "build" its state

#

Because that's all LLMs do, they build their context one message at a time

inland birch
#

So printing an ID digest could help